Quotas and network
How much compute you can run at once, how long it can live, and what network traffic sandboxes and VMs may send and receive.
Trust tiers
Each workspace has a trust tier: T0, T1 or T2. The Settings page shows it as "Egress trust tier". New workspaces start at T0. Tiers are resource quotas against abuse, not spending limits. A paid Yi Labs plan (Pro or above) moves you to T1 automatically. For T2, use the contact form.
The default quotas:
| T0 | T1 | T2 | |
|---|---|---|---|
| Sandboxes and VMs at once | 1 | 2 | 8 |
| vCPUs at once | 2 | 8 | 32 |
| Longest sandbox lifetime or VM lease | 1 hour | 24 hours | 7 days |
- Sandboxes and VMs count together. Every one that isn't deleted counts, stopped VMs included.
- The lifetime limit applies to a sandbox's
timeout_s, a VM'shours, and how far ahead a VM's lease may reach after a renewal. - We can set different quotas for your workspace. Ask us.
A request over the quota gets 409 quota_exceeded, with "Compute quota reached", "Requested lifetime exceeds the organization quota" or "Renewal exceeds the maximum lease".
The tier and the network rules are fixed for each resource when it's created. A tier change applies to resources created after it.
Outbound network
Sandboxes and VMs may connect out to:
- the internet on TCP ports 80 (HTTP) and 443 (HTTPS),
- the platform's DNS.
These are always blocked, whatever the tier:
- mail ports 25, 465 and 587,
- private address ranges (
10.0.0.0/8,172.16.0.0/12,192.168.0.0/16), so resources can't reach each other, - destinations on our threat lists.
Everything else is blocked.
Inbound network
- Sandboxes: nothing. A sandbox has no public address, and you reach it only through the API.
- VMs: TCP port 22 (SSH) only, and only from the ranges in
allowed_cidrs. IPv4 ranges must be/8or narrower, IPv6 ranges/32or narrower. Everything else is blocked.
Other limits
| Limit | Value |
|---|---|
Sandbox lifetime (timeout_s) | 60 seconds to 7 days, within the tier |
Sandbox idle time (idle_timeout_s) | 60 seconds to 24 hours |
VM lease (hours) | 1 to 168 hours per request, within the tier |
| Time for a sandbox or VM to become ready | 10 minutes |
| A stopped VM is kept | 24 hours |
| SSH keys per VM | 10 |
| Address ranges per VM | 20 |
| Commands at once per sandbox | 4 |
| One command | 1 hour |
API request limits are in the API reference, model limits in Models API.